slowest-tests
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from BuildKite job logs, which introduces a potential surface for indirect prompt injection if logs contain adversarial content.
- Ingestion points: BuildKite raw logs are retrieved via
raw_log_urlas described inSKILL.md. - Boundary markers: The skill does not define clear boundaries or isolation constraints to prevent the agent from interpreting instructions that might be embedded in the log text.
- Capability inventory: The skill executes shell commands (
bun run,gh,bk) and uses Slack interaction tools (slack_send_message,slack_search_channels). - Sanitization: The script
scripts/ci-slowest-tests.tsfilters and normalizes timing data headers, but the final text output is handled by the agent to compose Slack messages and threads, creating an exposure channel.
Audit Metadata